Open Models Almanac
Methodsinference & servingreasoning control

specific method · filed under inference & serving

Max thinking mode with self-set compute budget

Max thinking mode is a reasoning-control setting that lets the model itself determine the right thinking/test-time compute budget for a given problem.1 It replaces user-set effort levels with an automatic choice between off and max, and can sustain coherent, productive thinking over long stretches.2,3

Trade-off Max thinking substantially raises benchmark scores, but users lose fine-grained low-medium-high effort control.4,2

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. two thinking modes: off and max (enabled by default) where it determines the right thinking/test-time compute budget for a given problem — Introducing Laguna S 2.1
  2. We are releasing this model without user-configurable effort (low-medium-high) control — Introducing Laguna S 2.1
  3. We have observed coherent, productive thinking over several hours and hundreds of thousands of tokens in length — Introducing Laguna S 2.1
  4. Max thinking lifts S 2.1’s score on Terminal-Bench 2.1 from 60.4% to 70.2 — Introducing Laguna S 2.1

Also called max thinking, Thinking modes (off and max), two modes: off and max.

sources
2
model
1
lab adopt it
1
strongest
default

Documented in

Poolside

Evidence

3 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
used 1default 2

Laguna-S-2.1

Poolsideadopted2 sources

In max mode the model sets its own test-time compute budget.

defaultinference servingin Laguna S 2.1Poolside
2 more from this model

Laguna S 2.1 has two modes: off and max, with max enabled by default.

usedinference servingin Laguna S 2.1Poolside

Laguna S 2.1 has two thinking modes: off and max (enabled by default) where it determines the right thinking/test-time compute budget for a given problem.

defaultinference servingin Laguna S 2.1Poolside
↑